Just going from the OfGen slides, clearly those that are supposed to understand these things consider there is a potential issue. That we've not had such an event in the past doesn't mean that one can't occur in the future - and if RoCoF is a trigger for large amounts of embedded generation to disconnect, then it's certainly an event to consider (even if after consideration it's deemed "unlikely" to occur).
AIUI the main concern isn't loss of supply to the one premises - I imagine it's relatively easy to work out if the inverter is suddenly not connected to anything. The main concern is islanding - where a block of the distribution network gets disconnected from the grid and the embedded generation continues supplying it.
While the 89kW water turbine I mentioned earlier could supply a large area, the relatively poor control of frequency (low inertia, mechanical control of input power) would make it fairly easy to detect and trip on RoCoF. I suspect voltage variations would be quite significant as well given the large power output, and the low likelihood of load being close to output - giving another trigger for tripping.
As an aside to that, part of the control system includes a diverter flap which can quickly be flipped into position between the nozzle and the turbine wheel. Primary control is via a spear valve - which is limited in speed of movement, and more critically must be limited in speed to avoid hydraulic shock (slam it shut, inertia in the flowing water causes a pressure spike, "something" bursts). That has happened on a commercial size hydro plant - during maintenance a rock got left in the feed pipes, at some point it went down the pipe, hit the spear valve with sufficient force to break the tip off, the tip then jammed in the nozzle, and the feed pipes behind the turbine hall exploded due to the pressure spike
I imagine that would have been a sight to see.
No I don't have a reference, I recall photos shown during a talk a few years ago - a search brings up several other cases of pipe or penstock failure, but not the one I had in mind.
What I don't know about is the crystal oscillator stable, digitally controlled, solar PV inverters. I imagine the response from these could be interesting in an islanded situation.
Does the frequency suddenly increase as they all try to lead the supply, or does the frequency remain stable and the voltage suddenly take off ? Do they all shut down, or could some remain online when the others have shutdown and the voltage returns to normalish when enough have shutdown that supply is close to load ? Is it possible (given a big enough island, and hence diversity) for the supply to remain stable enough that some inverters come back online ? Is there still enough simple* load in a typical distribution area for the voltage-power curve to still show a positive slope, or is it now (with the amount of SMPS load) getting to the point where it's negative and so any system relying on load varying with voltage to help with balancing becomes unstable ?
* Simple, as in there's a positive (and proportional) link between supply voltage and load power.
A crisis which would not arise if those responsible for providing generating capacity had not decided that a spiffing way to provide it would be millions of tiny generators tied into the grid.
I don't think that millions of tiny generators is the problem. The proportion, and characteristics, of distributed generation capacity is the problem, and I recall having read many years ago (long before all the subsidies made the issue mushroom as it has done) discussion related to the amount of distributed generation that the system could have and still remain stable.